Household garbage incineration flue gas dioxin emission control method based on big data algorithm
Through big data algorithms and random forest models, dioxin emissions in domestic waste incineration flue gas are monitored and adjusted in real time, which solves the problem that the existing technology cannot obtain dioxin emission conditions and provide operating conditions adjustment guidelines in real time, and achieves the correct Effective control of dioxin emissions.
Patent Information
- Application Number
- CN202310646896.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-06-02
- Publication Date
- 2024-12-03
AI Technical Summary
The prior art cannot obtain the dioxin emissions in domestic waste incineration flue gas in real time, and cannot provide operating personnel with operating conditions adjustment guidance, resulting in the inability to effectively control dioxin emissions.
Using a method based on big data algorithm, a dioxin prediction model is constructed by constructing a database, using a stochastic forest algorithm, and combining incineration working condition indicators, a dioxin early warning model is formed, and the working condition indicators are monitored and adjusted in real time to control dioxin emissions. .
Real-time monitoring and working condition adjustment of dioxin emissions are achieved, specific control guidelines are provided, and the control effect of dioxin emissions is improved. It is suitable for a wide range of domestic waste incineration power plants.
Smart Images

Figure CN119063003A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of domestic waste incineration, and in particular to a method for controlling dioxin emissions from domestic waste incineration flue gas based on a big data algorithm. Background Art
[0002] Dioxin emissions from the municipal solid waste incineration industry have always been a widely-concerned issue for the public, and are also an important factor causing the "NIMBY effect" of municipal solid waste incineration power plants. Dioxin emissions from municipal solid waste incineration facilities have characteristics such as occasional mutations, and are affected by many factors. On the one hand, they are affected by the composition of municipal solid waste, and on the other hand, they are mainly affected by the incineration conditions of the incinerator (furnace temperature, furnace turbulence, flue gas residence time, etc.);
[0003] 1. Currently, among the existing technologies, the manual monitoring technology for dioxins has a long cycle and cannot obtain the dioxin emission concentration in real time in actual production;
[0004] 2. Although online monitoring and early warning technology can obtain the emission status of dioxins to a certain extent, it cannot provide operators with guidance on actual working condition adjustments;
[0005] 3. The method of monitoring precursors such as chlorobenzene is, on the one hand, a technology that is still under verification, and on the other hand, similar to online monitoring and early warning technology, it is unable to provide guidance to operators on actual operating condition adjustments. To address the above problems, the inventors proposed a method for controlling dioxin emissions from domestic waste incineration flue gas based on big data algorithms to solve the above problems. Summary of the invention
[0006] In order to solve the problem that the existing technology cannot simultaneously obtain the real-time status of dioxin emissions and guide operators to adjust operating indicators; the purpose of the present invention is to provide a method for controlling dioxin emissions from domestic waste incineration flue gas based on big data algorithms.
[0007] In order to solve the above technical problems, the present invention adopts the following technical solution: a method for controlling dioxin emissions from flue gas from the incineration of domestic waste based on a big data algorithm, comprising the following steps:
[0008] S1. Build a database, form a database module, collect manual monitoring data of dioxins from several waste incineration power plants and conventional indicators of the flue gas emission continuous monitoring system CEMS at the corresponding time;
[0009] S2. Random forest algorithm construction, training and testing, and establishment of a dioxin prediction model based on random forest;
[0010] S3. Use the dioxin prediction model to predict the effective historical CEMS routine indicators of an incineration power plant;
[0011] S4. Construct a dataset of CEMS routine indicators, incineration condition indicators and dioxin prediction results for a municipal solid waste incineration power plant;
[0012] S5. Reconstruct, train and test the dioxin prediction model based on CEMS conventional indicators, incineration condition indicators and dioxin prediction results to form a new dioxin prediction and early warning model that includes incineration condition indicators;
[0013] S6. Obtain the weight index of the dioxin prediction and early warning module, divide the top three weighted indexes into those that exceed the standard and those that do not exceed the standard, obtain the range of each index, and input it into the incineration condition index early warning module;
[0014] S7, inputting the real-time indicator monitoring data and dioxin prediction results into the incineration condition indicator early warning module;
[0015] S8. The incineration condition indicator warning module proposes control requirements for each indicator based on the current incineration condition indicator values and dioxin prediction results, and displays them to the operating personnel.
[0016] Preferably, in S1, the collected parameters include the following:
[0017] The collected manual monitoring data of dioxins were used as data labels, and the hourly average values of sulfur dioxide, nitrogen oxides, particulate matter, hydrogen chloride, carbon monoxide, incinerator type, and incineration scale in the corresponding period were used as features to construct a data set;
[0018]
[0019] in:
[0020]
[0021] Where a i The features are specifically the i-th data of sulfur dioxide, nitrogen oxides, particulate matter, hydrogen chloride, and carbon monoxide (five parameters), li refers to the label of the manual monitoring result of dioxins (exceeding the standard, not exceeding the standard), and k refers to the sample size of the data set.
[0022] Preferably, in S2, random forest algorithm construction, training and testing include: constructing a random forest model using the Gini index or out-of-bag error, and dividing the data collected in S1 into a training set and a test set in an 8:2 ratio.
[0023] Preferably, a preferred embodiment of S2 is as follows:
[0024] Use the Gini index or out-of-bag error to select the optimal feature and determine the optimal binary split point for the feature;
[0025] SS1, for feature a i , for every possible value vi , according to the manual monitoring result label l = exceeding / not exceeding the standard for dioxin of sample point a = v, D1 is divided into 2 parts and its Gini index is calculated;
[0026] SS2, select the feature with the smallest Gini index and its value as the optimal feature and optimal split point, and generate child nodes based on the optimal feature and optimal split point;
[0027] SS3, repeat steps SS1 and SS2 until the number of samples in the node is less than the predetermined threshold, or the Gini index of the sample set is less than the predetermined threshold (the samples basically belong to the same class), or there are no more features;
[0028] SS4. Generate a random forest and make predictions on the test set.
[0029] Preferably, in S3, a dioxin prediction model is used to predict the effective historical CEMS conventional indicators of an incineration power plant, that is, all data of a certain waste power plant are input into the model prediction, and the prediction result is obtained.
[0030] Preferably, in S4, a data set of CEMS conventional indicators, incineration operating condition indicators and dioxin prediction results of a municipal solid waste incineration power plant is constructed, that is, the CEMS conventional indicators, incineration operating condition indicators and the dioxin results predicted in S3 are reconstructed into a new data set D2;
[0031]
[0032] In S4, the incineration operating condition indicators refer to flue gas volume, oxygen content, humidity, flue gas temperature, flue gas pressure, average furnace temperature, DCS temperature, upper, middle and lower layer temperatures, main steam flow, economizer outlet oxygen content, first-stage economizer inlet flue gas temperature, economizer outlet temperature, primary air supply volume, secondary air supply volume, bag inlet temperature, bag outlet temperature, bag pressure difference, activated carbon injection weight, and environmental protection material usage.
[0033] Preferably, in S5, the dioxin prediction and warning model including the operating condition data is reconstructed, trained and tested, the data set is divided into a test set and a training set at a ratio of 8:2, the random forest model is reconstructed, and a new dioxin prediction model is obtained.
[0034] Preferably, a preferred method of S6 is specifically:
[0035] SS1. According to the weight of random forest, the top three indicators are used as key control indicators;
[0036] SS2. According to the prediction results of the new dioxin prediction model, the CEMS conventional indicators and incineration condition indicators are grouped and summarized, and the frequency histogram is drawn to visualize the group statistical results;
[0037] SS3. According to the statistical distribution results of step SS2, the distribution range of key parameters is analyzed, and the optimal control range of key parameters is obtained with more than 80% data separation as the boundary.
[0038] Compared with the prior art, the present invention has the following beneficial effects:
[0039] 1. In the present invention, the early warning and guidance range of the incineration operating condition indicators are proposed according to the weights and statistical results of the incineration operating condition indicators, so as to solve the problem that it is impossible to simultaneously obtain the real-time status of dioxin emissions and guide the operating personnel to adjust the operating condition indicators according to the real-time status;
[0040] 2. In the present invention, the data modeling is based on the domestic waste incineration power plants across the country, and the model has a wider range of adaptability and stronger generalization ability;
[0041] 3. In the present invention, the reconstruction model of the incineration condition index is combined to realize the detection and early warning of dioxins and provide specific parameter control guidance to the on-site operating personnel. BRIEF DESCRIPTION OF THE DRAWINGS
[0042] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0043] Figure 1 is a flow chart of an embodiment of the present invention. DETAILED DESCRIPTION
[0044] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0045] like Figure 1 As shown, the present invention provides a method for controlling dioxin emissions from flue gas from domestic waste incineration based on a big data algorithm, comprising the following steps:
[0046] S1. Build a database, form a database module, collect manual monitoring data of dioxins from several waste incineration power plants and conventional indicators of the flue gas emission continuous monitoring system CEMS at the corresponding time;
[0047] S2. Random forest algorithm construction, training and testing, and establishment of a dioxin prediction model based on random forest;
[0048] S3. Use the dioxin prediction model to predict the effective historical CEMS routine indicators of an incineration power plant;
[0049] S4. Construct a dataset of CEMS routine indicators, incineration condition indicators and dioxin prediction results for a municipal solid waste incineration power plant;
[0050] S5. Reconstruct, train and test the dioxin prediction model based on CEMS conventional indicators, incineration condition indicators and dioxin prediction results to form a new dioxin prediction and early warning model that includes incineration condition indicators;
[0051] S6. Obtain the weight index of the dioxin prediction and early warning module, divide the top three weighted indexes into those that exceed the standard and those that do not exceed the standard, obtain the range of each index, and input it into the incineration condition index early warning module;
[0052] S7, inputting the real-time indicator monitoring data and dioxin prediction results into the incineration condition indicator early warning module;
[0053] S8. The incineration condition indicator warning module proposes control requirements for each indicator based on the current incineration condition indicator values and dioxin prediction results, and displays them to the operating personnel.
[0054] In S1, the collected parameters include the following:
[0055] The collected manual monitoring data of dioxins were used as data labels, and the hourly average values of sulfur dioxide, nitrogen oxides, particulate matter, hydrogen chloride, carbon monoxide, incinerator type, and incineration scale in the corresponding period were used as features to construct a data set;
[0056]
[0057] in:
[0058]
[0059] Where a i The features are specifically the i-th data of sulfur dioxide, nitrogen oxides, particulate matter, hydrogen chloride, and carbon monoxide (five parameters), li refers to the label of the manual monitoring result of dioxins (exceeding the standard, not exceeding the standard), and k refers to the sample size of the data set.
[0060] In S2, the random forest algorithm construction, training and testing include: constructing a random forest model using the Gini index or out-of-bag error, and dividing the data collected in S1 into training and test sets in an 8:2 ratio.
[0061] A preferred method of S2 is as follows:
[0062] Use the Gini index or out-of-bag error to select the optimal feature and determine the optimal binary split point for the feature;
[0063] SS1, for feature a i , for every possible value v i , according to the manual monitoring result label l = exceeding / not exceeding the standard for dioxin of sample point a = v, D1 is divided into 2 parts and its Gini index is calculated;
[0064] SS2, select the feature with the smallest Gini index and its value as the optimal feature and optimal split point, and generate child nodes based on the optimal feature and optimal split point;
[0065] SS3, repeat steps SS1 and SS2 until the number of samples in the node is less than the predetermined threshold, or the Gini index of the sample set is less than the predetermined threshold (the samples basically belong to the same class), or there are no more features;
[0066] SS4. Generate a random forest and make predictions on the test set.
[0067] In S3, the dioxin prediction model is used to predict the effective historical CEMS conventional indicators of an incineration power plant, that is, all the data of a certain waste power plant are input into the model prediction and the prediction results are obtained.
[0068] In S4, a dataset of CEMS conventional indicators, incineration operating condition indicators and dioxin prediction results of a municipal solid waste incineration power plant is constructed, that is, a new dataset D2 is reconstructed by combining CEMS conventional indicators, incineration operating condition indicators and dioxin results predicted by S3;
[0069]
[0070] In S4, the incineration operating condition indicators refer to flue gas volume, oxygen content, humidity, flue gas temperature, flue gas pressure, average furnace temperature, DCS temperature, upper, middle and lower layer temperatures, main steam flow, economizer outlet oxygen content, first-stage economizer inlet flue gas temperature, economizer outlet temperature, primary air supply volume, secondary air supply volume, bag inlet temperature, bag outlet temperature, bag pressure difference, activated carbon injection weight, and environmental protection material usage.
[0071] In S5, the dioxin prediction and warning model containing the operating condition data is rebuilt, trained and tested, the data set is divided into a test set and a training set at a ratio of 8:2, the random forest model is rebuilt, and a new dioxin prediction model is obtained.
[0072] A preferred method of S6 is as follows:
[0073] SS1. According to the weight of random forest, the top three indicators are used as key control indicators;
[0074] SS2. According to the prediction results of the new dioxin prediction model, the CEMS conventional indicators and incineration condition indicators are grouped and summarized, and the frequency histogram is drawn to visualize the group statistical results;
[0075] SS3. According to the statistical distribution results of step SS2, the distribution range of key parameters is analyzed, and the optimal control range of key parameters is obtained with more than 80% data separation as the boundary.
[0076] An embodiment of the above steps:
[0077] Collect manual dioxin monitoring data from several domestic waste incineration power plants and CEMS routine indicators for the corresponding time periods, and establish a dioxin emission early warning prediction model based on random forests;
[0078] About 20,000 data were collected from two incinerators in a municipal solid waste incineration power plant in several time periods, including characteristic parameters: measured value of particulate matter, converted value of particulate matter, measured value of sulfur dioxide, converted value of sulfur dioxide, measured value of nitrogen oxides, converted value of nitrogen oxides, measured value of hydrogen chloride, converted value of hydrogen chloride, measured value of carbon monoxide, converted value of carbon monoxide, flue gas oxygen content, exhaust gas temperature, flue gas humidity, flue gas pressure, flue gas flow, furnace average temperature, DCS temperature, main steam flow, economizer outlet oxygen content, first-stage economizer inlet flue gas temperature, economizer outlet temperature, primary air supply volume, secondary air supply volume, bag inlet temperature, bag outlet temperature, bag pressure difference, activated carbon weighing weight. After eliminating the start-up and shutdown data and missing data, 11,143 valid data were obtained;
[0079] Input the particulate matter conversion value, sulfur dioxide conversion value, nitrogen oxide conversion value, hydrogen chloride conversion value, and carbon monoxide conversion value into the dioxin prediction model to obtain the dioxin prediction results;
[0080] Select 22 parameters including particulate matter conversion value, sulfur dioxide conversion value, nitrogen oxide conversion value, hydrogen chloride conversion value, carbon monoxide conversion value, flue gas oxygen content, exhaust temperature, flue gas humidity, flue gas pressure, flue gas flow rate, furnace average temperature, DCS temperature, main steam flow rate, economizer outlet oxygen content, primary economizer inlet flue gas temperature, economizer outlet temperature, primary air supply volume, secondary air supply volume, bag inlet temperature, bag outlet temperature, bag pressure difference, activated carbon weighing weight as features and the above model prediction results as labels to retrain the random forest model;
[0081] The model parameter results and the weights of each parameter are shown in Table 1 and Table 2:
[0082] precision recall f1-score support Meet the Standard 0.99 0.93 0.96 1919 Exceeding the standard 0.70 0.95 0.80 310 accuracy 0.94 2229 macro avg 0.84 0.94 0.88 2229 weighted avg 0.95 0.94 0.94 2229
[0083] Table 1 Dioxin model reconstruction test set results
[0084]
[0085]
[0086] Table 2 Weight table
[0087] The distribution of the top three feature data in the case of exceeding the standard and not exceeding the standard is counted respectively, and the 10th and 90th of the features in the case of meeting the standard are counted as the target control range. The results are shown in Table 3;
[0088] Sulfur dioxide conversion value Flue gas humidity Carbon monoxide conversion value Meet the Standard 11.7~43.2 14.5~25.3 1.7~20.1 Exceeding the standard 3.0~8.7 5.3~21.9 0.1~11.8
[0089] Table 3 Optimal control range of incineration operating conditions
[0090] When real-time data show that the dioxin prediction result exceeds the standard, the target control range is the standard range for the three incineration condition indicators: sulfur dioxide conversion value, flue gas humidity, and carbon monoxide conversion value.
[0091] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.
Claims
1. A method for controlling dioxin emissions from flue gas from domestic waste incineration based on big data algorithms, characterized in that: The following steps are involved: S1. Build a database, form a database module, collect manual monitoring data of dioxins from several waste incineration power plants and conventional indicators of the flue gas emission continuous monitoring system CEMS at the corresponding time; S2. Random forest algorithm construction, training and testing, and establishment of a dioxin prediction model based on random forest; S3. Use the dioxin prediction model to predict the effective historical CEMS routine indicators of an incineration power plant; S4. Construct a dataset of CEMS routine indicators, incineration condition indicators and dioxin prediction results for a municipal solid waste incineration power plant; S5. Reconstruct, train and test the dioxin prediction model based on CEMS conventional indicators, incineration condition indicators and dioxin prediction results to form a new dioxin prediction and early warning model that includes incineration condition indicators; S6. Obtain the weight index of the dioxin prediction and early warning module, divide the top three weighted indexes into those that exceed the standard and those that do not exceed the standard, obtain the range of each index, and input it into the incineration condition index early warning module; S7, inputting the real-time indicator monitoring data and dioxin prediction results into the incineration condition indicator early warning module; S8. The incineration condition indicator warning module proposes control requirements for each indicator based on the current incineration condition indicator values and dioxin prediction results, and displays them to the operating personnel.
2. The method for controlling dioxin emissions from domestic waste incineration flue gas based on big data algorithm as claimed in claim 1, characterized in that: In S1, the collected parameters include the following: The collected manual monitoring data of dioxins were used as data labels, and the hourly average values of sulfur dioxide, nitrogen oxides, particulate matter, hydrogen chloride, carbon monoxide, incinerator type, and incineration scale in the corresponding period were used as features to construct a data set; in: Where a i The features are specifically the i-th data of sulfur dioxide, nitrogen oxides, particulate matter, hydrogen chloride, and carbon monoxide (five parameters), li refers to the label of the manual monitoring result of dioxins (exceeding the standard, not exceeding the standard), and k refers to the sample size of the data set.
3. The method for controlling dioxin emissions from domestic waste incineration flue gas based on big data algorithm as claimed in claim 1, characterized in that: In S2, the random forest algorithm construction, training and testing include: constructing a random forest model using the Gini index or out-of-bag error, and dividing the data collected in S1 into training and test sets in an 8:2 ratio.
4. The method for controlling dioxin emissions from domestic waste incineration flue gas based on big data algorithm as claimed in claim 1, characterized in that: A preferred method of S2 is as follows: Use the Gini index or out-of-bag error to select the optimal feature and determine the optimal binary split point for the feature; SS1, for feature a i , for every possible value v i , according to the manual monitoring result label l = exceeding / not exceeding the standard for dioxin of sample point a = v, D1 is divided into 2 parts and its Gini index is calculated; SS2, select the feature with the smallest Gini index and its value as the optimal feature and optimal split point, and generate child nodes based on the optimal feature and optimal split point; SS3, repeat steps SS1 and SS2 until the number of samples in the node is less than the predetermined threshold, or the Gini index of the sample set is less than the predetermined threshold (the samples basically belong to the same class), or there are no more features; SS4. Generate a random forest and make predictions on the test set.
5. The method for controlling dioxin emissions from domestic waste incineration flue gas based on big data algorithm as claimed in claim 1, characterized in that: In S3, the dioxin prediction model is used to predict the effective historical CEMS conventional indicators of an incineration power plant, that is, all the data of a certain waste power plant are input into the model prediction and the prediction results are obtained.
6. The method for controlling dioxin emissions from domestic waste incineration flue gas based on big data algorithm as claimed in claim 1, characterized in that: In S4, a dataset of CEMS conventional indicators, incineration operating condition indicators and dioxin prediction results of a municipal solid waste incineration power plant is constructed, that is, a new dataset D2 is reconstructed by combining CEMS conventional indicators, incineration operating condition indicators and dioxin results predicted by S3; In S4, the incineration operating condition indicators refer to flue gas volume, oxygen content, humidity, flue gas temperature, flue gas pressure, average furnace temperature, DCS temperature, upper, middle and lower layer temperatures, main steam flow, economizer outlet oxygen content, first-stage economizer inlet flue gas temperature, economizer outlet temperature, primary air supply volume, secondary air supply volume, bag inlet temperature, bag outlet temperature, bag pressure difference, activated carbon injection weight, and environmental protection material usage.
7. The method for controlling dioxin emissions from domestic waste incineration flue gas based on big data algorithm as claimed in claim 1, characterized in that: In S5, the dioxin prediction and warning model containing the operating condition data is rebuilt, trained and tested, the data set is divided into a test set and a training set at a ratio of 8:2, the random forest model is rebuilt, and a new dioxin prediction model is obtained.
8. The method for controlling dioxin emissions from flue gas from domestic waste incineration based on big data algorithm as claimed in claim 1, characterized in that: A preferred method of S6 is as follows: SS1. According to the weight of random forest, the top three indicators are used as key control indicators; SS2. According to the prediction results of the new dioxin prediction model, the CEMS conventional indicators and incineration condition indicators are grouped and summarized, and the frequency histogram is drawn to visualize the group statistical results; SS3. According to the statistical distribution results of step SS2, the distribution range of key parameters is analyzed, and the optimal control range of key parameters is obtained with more than 80% data separation as the boundary.
